Disrupted Iron Homeostasis Causes Dopaminergic Neurodegeneration in Mice

Disrupted iron homeostasis causes dopaminergic neurodegeneration in mice

Pavle Mataka, Andrija Mataka, Sarah Moustafaa, Dipendra K. Aryalb,c,d,e, Eric J. Bennerf, William Wetselb,c,d,e, and Nancy C. Andrewsa,f,1

aDepartment of Pharmacology and Cancer Biology, Duke University School of Medicine, Durham, NC 27705; bDepartment of Psychiatry and Behavioral Sciences, Duke University School of Medicine, Durham, NC 27705; cDepartment of Cell Biology, Duke University School of Medicine, Durham, NC 27705; dDepartment of Neurobiology, Duke University School of Medicine, Durham, NC 27705; eMouse Behavioral and Neuroendocrine Analysis Core Facility, Duke University School of Medicine, Durham, NC 27705; and fDepartment of Pediatrics, Duke University School of Medicine, Durham, NC 27705

This contribution is part of the special series of Inaugural Articles by members of the National Academy of Sciences elected in 2015.

Contributed by Nancy C. Andrews, January 27, 2016 (sent for review September 30, 2015; reviewed by Michael K. Georgieff and Barry H. Paw) Disrupted brain iron homeostasis is a common feature of neuro- offers a very sensitive approach to detect cellular iron deficiency degenerative disease. To begin to understand how neuronal iron and surfeit (12). handling might be involved, we focused on dopaminergic neurons Iron homeostasis is altered locally, in the affected part of the and asked how inactivation of transport proteins affected iron brain, in most human neurodegenerative disorders (13). Iron ac- homeostasis in vivo in mice. Loss of the cellular iron exporter, cumulates in the substantia nigra (SN) in Parkinson’s disease (PD) ferroportin, had no apparent consequences. However, loss of (14–16), although it is unsettled whether increased iron is in neu- transferrin receptor 1, involved in iron uptake, caused neuronal ropil (17) or dopaminergic (DA) neurons (18), or both. Excess DA iron deficiency, age-progressive degeneration of a subset of dopami- neuron iron has been proposed to contribute to disease pathogen- nergic neurons, and motor deficits. There was gradual depletion of esis (19). Conversely, several reports suggest that systemic iron dopaminergic projections in the striatum followed by death of deficiency, rather than iron overload, may predispose to PD. PD has dopaminergic neurons in the substantia nigra. Damaged mitochon- been associated with a history of anemia years before the onset of dria accumulated, and gene expression signatures indicated attempted motor symptoms (20), with multiple blood donations that deplete axonal regeneration, a metabolic switch to glycolysis, oxidative stress, iron stores (21), and with low serum iron (22). Genetic pre- and the unfolded protein response. We demonstrate that loss of dispositiontoironoverloadappearstobeprotective(23–25). Thus, transferrin receptor 1, but not loss of ferroportin, can cause neuro- iron deposition in the SN may or may not be in DA neurons degeneration in a subset of dopaminergic neurons in mice. themselves, and iron overload, iron deficiency, or both might be iron | transferrin receptor | ferroportin | dopaminergic neuron | deleterious to DA neurons. neurodegeneration We used conditional KO mice to study how perturbations of iron transport affect DA neuron survival and function. Others previously suggested that inactivation of FPN results in iron accumu- n the brain, iron is needed for mitochondrial respiration and lation in DA neurons, causing PD (19). However, it was not clear Isynthesis of myelin, neurotransmitters, and monoamine oxidases (1, 2). Transferrin receptor 1 (TFR1) is known to be required for iron uptake by some, but not all, cell types (3–7). Its ligand, trans- Significance ferrin (TF), carries extracellular iron. Fe2-TF/TFR1 is internalized by receptor-mediated endocytosis and trafficked to endosomes, The brain requires iron for mitochondrial respiration and synthesis where low pH liberates iron from TF, allowing it to leave the of myelin, neurotransmitters, and monoamine oxidases. Iron ac- endosome through transmembrane transport. This transport is cumulates in distinct parts of the brain in patients with neurode- mediated by divalent metal transporter 1 (DMT1/SLC11A2) in generative diseases, and some have proposed that neurons die erythroblasts and possibly other cell types (8). Imported iron is used because they contain too much iron. Neuronal iron handling is not directly, incorporated into heme or Fe-S clusters, or stored in fer- well understood. We focused on dopaminergic neurons, affected ritin. It has been widely assumed that neuronal iron uptake depends in Parkinson’s disease, and manipulated molecules involve in iron uptake and release. We showed that loss of ferroportin, which on TFR1-mediated endocytosis of Fe2-TF (9), but in vivo evidence is scant. Expression of a dominant negative form of Tfr1 was shown exports cellular iron, had no apparent effect. In contrast, loss of to decrease iron uptake by hippocampal CA1 pyramidal neurons transferrin receptor, involved in iron uptake, caused neuronal iron (10). However, others noted that iron distribution does not corre- deficiency and neurodegeneration with features similar to late with Tfr1 expression in the brain and suggested that Tfr1 may Parkinson’s disease. We propose that neuronal iron deficiency have roles unrelated to iron uptake (11). may contribute to neurodegeneration in human disease. There is only one known mechanism for cellular iron release, involving the transmembrane transporter ferroportin (FPN). FPN Author contributions: P.M., E.J.B., W.W., and N.C.A. designed research; P.M., A.M., S.M., and D.K.A. performed research; W.W. contributed new reagents/analytic tools; P.M., A.M., D.K.A., exports iron, in collaboration with ceruloplasmin or a related E.J.B., W.W., and N.C.A. analyzed data; and P.M., W.W., and N.C.A. wrote the paper. ferroxidase, to be loaded onto TF (8). Intracellular iron homeo- Reviewers: M.K.G., University of Minnesota; and B.H.P., Brigham & Women’s Hospital, stasis is controlled by iron-regulatory proteins (IRPs), which rec- Harvard Medical School. ognize iron regulatory elements (IREs) in the untranslated The authors declare no conflict of interest. portions of mRNAs encoding proteins important for iron trans- Freely available online through the PNAS open access option. port or storage (12). In iron deficiency, IRPs bind IREs to block Data deposition: The microarray data reported in this paper have been deposited in the ribosomal entry onto ferritin mRNAs, precluding translation of Gene Expression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no. this iron storage protein, and IRPs stabilize mRNAs encoding GSE66730). TFR1 and SLC11A2, facilitating iron uptake. In iron surfeit, fer- See Commentary on page 3417. ritin mRNAs are actively translated to store excess iron and iron 1To whom correspondence should be addressed. Email: [email protected]. import-related mRNAs are degraded. This well-studied homeo- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. static mechanism allows cells to regulate their iron content and 1073/pnas.1519473113/-/DCSupplemental.

3428–3435 | PNAS | March 29, 2016 | vol. 113 | no. 13 www.pnas.org/cgi/doi/10.1073/pnas.1519473113 Downloaded by guest on October 2, 2021 why DA neurons would require active iron export or why com- (Fig. 1B). Neurophysiological and behavioral tests (29) showed no

pensatory changes in cellular iron import or storage would not differences at 16 mo. Th staining in the dorsal striatum and cell INAUGURAL ARTICLE maintain iron homeostasis. Accordingly, we found that mice lack- counts in the SN pars compacta (SNpc) were also indistinguish- ing Fpn in DA neurons had no apparent abnormalities in neuronal able at 19 mo (Fig. 1C). iron homeostasis and no evidence of neurodegeneration or DA Because the ventral midbrain contains multiple cell types, we neuron dysfunction. However, mice lacking Tfr1 in DA neurons could not use tissue to examine the iron content of DA neurons. developed a severe phenotype characterized by DA neuron iron We attempted to quantify iron content of DA neurons isolated insufficiency, progressive degeneration of neurons in the SN, de- from the ventral midbrain by FACS followed by inductively creased motor activity, and early death. We conclude that Tfr1 is coupled plasma MS (ICP-MS), but the amounts were too low. important for iron homeostasis in DA neurons of the SN and that Similarly, there was too little iron to visualize with Perls stain. As Fpn is dispensable. an alternative approach, we examined polysome-associated DA SEE COMMENTARY neuron mRNA for evidence of posttranscriptional regulation by Results the IRE/IRP system. Cellular iron surfeit should result in in- Deletion of Fpn in DA Neurons Has No Apparent Consequences. We creased Fth and Ftl and decreased Slc11a2 and Tfr1 mRNAs, but inactivated Fpn in DA neurons by crossing dopamine transporter we observed no significant differences between Fpn-CKO and (Dat)IRES Cre (26) and Fpnfl/fl mice (27) to obtain Fpn conditional CTR mice at 3 wk and 18 mo (Fig. 1D). We conclude that Fpn KO (Fpn-CKO) mice. We confirmed Fpn depletion by analyzing was not essential for DA neuron survival or function and Fpn- CKO DA neurons did not accumulate excess iron. polysome-associated mRNAs from ventral midbrain DA neurons. IRES To do this, we bred the floxed Fpn allele and Dat Cre transgene Tfr1 Is Critical for DA Neuron Iron Homeostasis, Function, and onto a Ribotag background (28) to allow for Cre-dependent ex- Survival. To determine whether Tfr1 was involved in DA neuron pression of hemagglutinin (HA)-tagged ribosomal protein L22. iron assimilation, we crossed Tfr1fl/fl and DatIRES Cre mice to We used the HA tag to immunoprecipitate polysome-associated obtain Tfr1-CKO animals. Ribotag polysome-associated mRNA mRNA from these mice at 3 wk and 18 mo of age and confirmed from ventral midbrain at 3 and 10 wk was highly enriched for DA that it was highly enriched for DA neuron transcripts and depleted neuron transcripts (Fig. S1 E–H) and depleted of Tfr1 mRNA of an oligodendrocyte marker (Fig. S1 A–D). By quantitative RT- (Fig. 2A). Tfr1-CKO mice were grossly indistinguishable from PCR (qPCR), we observed a marked decrease in polysome-asso- CTR littermates until 7 wk of age when they began to lose weight ciated Fpn mRNA at 3 wk of age (Fig. 1A), indicating that most (Fig. 2B) and died by wk 12 (Fig. 2C). Polysome-associated Fpn alleles had been inactivated. Residual Fpn mRNA may have mRNA at 3 and 10 wk (Fig. 2D)revealeddecreasedFth and Ftl MEDICAL SCIENCES come from contamination by other cell types or incomplete in- mRNAs and increased Tfr1 (primers spanning retained exons 16 activation of the floxed Fpn allele. and 17) and Slc11a2+IRE mRNAs in Tfr1-CKO ventral midbrain Fpn-CKO mice appeared similar to control (CTR) mice. There DA neurons, indicating that they were iron deficient. were no significant differences in basal activity in an open field At 10 wk, Tfr1-CKO mice had hunched posture, low pelvis, (OF) at 4, 14, or 18 mo or response to metamphetamine at 18 mo attenuated response to visual stimuli, and reduced reflexive

Fig. 1. Deletion of Fpn in DA neurons. White bars, control (CTR); black bars, Fpn-CKO. (A)QPCRanalysis of Fpn after immunoprecipitation of HA-tagged polysomes from CTR;Ribotag and Fpn-CKO;Ribotag ventral midbrain homogenates, relative to CTR = 1 (CTR n = 4; Fpn-CKO n = 5; mice 3 wk of age); Stu- dent’s t test, **P < 0.01. (B) Locomotor activity in OF. (Left) Basal cumulative distance traveled in the same cohort of mice at 4, 14, and 18 mo of age (mos) (n ≥ 8 for both genotypes/all time points). Repeated measures analyses of variance (RMANOVA) found a sig-

nificant effect of age [F(2,28) = 27.452, P < 0.001] but not genotype; (Right)19momiceweretreated with 2 mg/kg i.p. metamphetamine 60 min post- acclimatization in the OF. RMANOVA revealed a sig-

nificant time effect [F(2,28) = 4.871, P < 0.015], but other comparisons were not significant. Bonferroni comparisons showed that, compared with base- line, all animals displayed increased activity to metamphetamine that was significant 2 h postinjection (P < 0.046). Data are represented as mean ± SEM distance traveled in 1-h blocks of time (n = 8 for both genotypes). (C) Anti-Th immunostaining of the dorsal striatum and SN at 19 mos. (Upper) Representative images from the dorsal striatum and quantification of terminal density; (Lower) representative images from SNpc. Th+ and Nissl+ cell numbers were determined using stereology. Ter- minal density was quantified using ImageJ (rsbweb. nih.gov/ij/). (Scale bars: 200 μm.) Student’s t test indicated no significant differences in terminal den- sities or cell numbers between genotypes. (CTR n = 5; Fpn-CKO n = 5.) (D) Relative polysome-associated mRNA for iron-related genes in DA neurons of CTR and Fpn-CKO mice at 3 wk and 18 mos of age. CTR mRNA expression was adjusted to 1. At 3 wk CTR n = 4, Fpn-CKO n = 5; at 18 mos CTR n = 4, Fpn-CKO n = 3. Student’s t test indicated no significant differences between genotypes. All data are presented as means ± SEM.

Matak et al. PNAS | March 29, 2016 | vol. 113 | no. 13 | 3429 Downloaded by guest on October 2, 2021 behavior (Fig. 2B and Table S1). Earlier they showed decreased Characterization of Tfr1-CKO DA Neurons. Because neurons require locomotor activity in an OF (Fig. 3A), age-progressive reduction iron for mitochondrial biogenesis (30) and respiration (31), we in steps taken with their front legs when their rear legs were examined mitochondrial morphology by electron microscopy, suspended (Fig. 3B), and delayed removal of forepaws placed on identifying DA neurons by immunogold localization of DAT. a bar (Fig. 3C). These latter behaviors are considered analogous Although SNpc mitochondria appeared normal in 6-wk Tfr1- to human akinesia and catalepsy, respectively. CKO mice, at 8 wk, they were enlarged with abnormal cristae At 7 wk, we observed a marked decrease in L-dihydroxy- (Fig. 5 A–F). In contrast, mitochondria in VTA DA neurons of phenylalanine (L-DOPA) synthesis rates (Fig. 3D) and tissue 8-wk Tfr1-CKO mice appeared similar to CTRs (Fig. S4). We dopamine levels (Fig. 3E) in the dorsal and ventral striata crossed Tfr1-CKO mice with mtYFP transgenic mice that ex- of Tfr1-CKO mice, although there was no significant difference press a mitochondria-targeted GFP in the presence of Cre in Th mRNA. Other brain regions were not affected. Dopa- recombinase (32). Tfr1-CKO mitochondria appeared fused or mine metabolites 3,4-dihydroxyphenylacetic acid (DOPAC) and aggregated in SNpc DA neurons (Fig. 5 G and H). Using ste- homovanillic acid (HVA) were also decreased in the ventral and reology at 8 wk, we counted GFP aggregates in SNpc DA neu- dorsal striata of Tfr1-CKO mice (Table S2). Tfr1-CKO and CTR rons. In CTR mice, 4 ± 0.03% of cells contained GFP aggregates mice were habituated to an OF for 1 h, treated with carbidopa, (n = 4 mice); in Tfr1-CKO mice, 73 ± 0.84% of cells contained and given L-DOPA 10 min later (Fig. 3F). After L-DOPA, GFP aggregates (n = 3 mice; P < 0.0001). We also observed we observed a marked increase in locomotion in Tfr1-CKO proteinaceous inclusions in SNpc DA dendrites of Tfr1-CKO but mice relative to CTRs. We treated with the mixed D1 and D2 not CTR mice (Fig. 5I). receptor agonist apomorphine and found that Tfr1-CKO animals We evaluated gene expression in ventral midbrain DA neu- were hyperresponsive (Fig. 3G). Tfr1-CKO mice had in- rons using Ribotag polysome-associated mRNA to probe Affy- creased locomotion in response to 2.5 and 5 mg/kg SKF-81297, a metrix microarrays. Data have been deposited online (NCBI D1 agonist (Fig. 3H), and a potentiated response to 0.5 mg/kg accession no. GSE66730). Pertinent results, shown as calcu- quinpirole, a D2 agonist (Fig. 3I). lated fold-change, are shown in Fig. 6 A–E. We validated gene Th staining in the dorsal striatum was similar in CTR and Tfr1- expression by qPCR in separate biological replicates (Figs. S5 CKO mice at 3 wk, but progressively decreased in Tfr1-CKO and S6). Genes with more than twofold change were subjected + animals thereafter (Fig. 4A). Th cell counts in the SNpc of Tfr1- to Gene Set Enrichment Analysis [(GSEA/MSigDB) (33, 34)]. CKO mice similarly decreased at 5, 7, and 10 wk (Fig. 4 B and C). GSEA revealed induction of several pathways previously asso- + SNpc Nissl neuron counts were similar at 3 and 5 wk, but sig- ciated with neurodegeneration (35)—hypoxia, mTorc1 signaling, nificantly decreased at 7 and 10 wk in Tfr1-CKO mice. We ob- the p53 pathway, the unfolded protein response (UPR), and — served later cell loss in the ventral tegmental area (VTA), despite glycolysis at both 3 and 10 wk. Genes associated with apoptosis efficient Cre expression there and in other DA neurons (Fig. S2 were induced at both ages, but more strikingly at 10 wk. A–D). Terminals and DA neurons in the nucleus accumbens and Chung et al. identified 52 genes that were expressed at higher olfactory bulb appeared to be relatively spared even at 10 wk (Fig. levels in SN neurons than VTA neurons (36). Of these, 19 were 4D). We stained for DOPA decarboxylase (Ddc) and obtained down-regulated in 10-wk Tfr1-CKO mice compared with con- similar results (Fig. S3). In summary, we observed a selective, trols: Anxa1, Aldh1a7, Sncg, Atp2a3, Srpx2, Tyrp1, Rerg, Grin2c, progressive loss of nigrostriatal DA projections in the dorsal Cd24a, Trhr, Sox6, Pvrl3, Lix1, Satb1, Igf1, Nrip3, Sox6, Vav3, and striatum and a later loss of DA neurons in the SNpc and then the Fgf1. No genes classified as specific to the SN were up-regulated VTA. Th staining in the ventral striatum was similar between in Tfr1-CKO mice, and no genes specific to the VTA were down- genotypes. These data are consistent with the behavioral abnor- regulated. Thus, gene expression was consistent with other ob- malities we observed in Tfr1-CKO mice. servations suggesting that VTA neurons were relatively spared. The most strongly up-regulated genes in Tfr1-CKO DA neurons were characteristic of axonal injury (37) and were similarly induced in rats after intrastriatal 6-hydroxydopamine treatment (38) (Fig. 6A). There was evidence of an attempted switch to glycolysis, with increased polysome-associated mRNAs encoding Pfkfb3 and glycolytic proteins Hk2, Slc2a1, and Ldha (Fig. 6B). Normal neurons preferentially use the pentose phosphate pathway, suppressing glycolysis through inhibition of Pfkfb3 (39). Increased Pfkfb3 activity and a switch to glycolysis contribute to neurodegeneration (40). Neurodegeneration is characterized by endoplasmic reticulum (ER) stress and induction of the UPR (41). We observed increased mRNAs encoding proteins involved in the protein kinase RNA-like endoplasmic reticulum (PERK) UPR pathway and proteins induced by PERK activation (Fig. 6C). Assessed by qPCR, there was increased Grp78/Bip, which triggers the UPR, as well as Atf4 and Herpud1,at3wk(Fig.6F). We also found evidence of the inositol- requiring enzyme 1 (IRE1) UPR pathway at 3 wk by qPCR: Xbp1 spliced isoform (Xbp1s) was increased without change in total Xbp1 mRNA (Fig. 6F). Finally, there was induction of Atf6, indicative of Fig. 2. DA neuron Tfr1 is critical for survival and iron homeostasis. the third (ATF6) UPR pathway (Fig. 6C). (A) Relative levels of Tfr1 mRNA in DA neurons (CTR n = 4, Tfr1-CKO n = 3); Polysome-associated mRNAs involved in glutathione synthesis ’ < unpaired Student s t test, **P 0.01. (B) Body weights over time and rep- were decreased (Fig. 6E), and heme oxygenase 1 (Hmox1)and resentative images at 12 wk. (C) Survival curve (CTR n = 6, Tfr1-CKO n = 7); P = 0.0007. (D) Relative polysome-associated mRNA for iron-related genes at sulfiredoxin 1 (Srxn1) were increased, indicative of oxidative stress 3 and 10 wk of age; Tfr1 sequences were from exons 16 and 17 which were (42). Prolonged UPR activation and oxidative stress can lead to not deleted. CTR mean adjusted to 1. At 3 wk CTR n = 4, Tfr1-CKO n = 3; at apoptosis, consistent with overrepresentation of Ddit3/Chop,a 10 wk CTR n = 3, Tfr1-CKO n = 4. Data are presented as means ± SEM; mediator of apoptotic death in SN DA neurons (43). We observed Student’s t test, *P < 0.05; **P < 0.01; ***P < 0.001. increased translation of mRNAs associated with p53 and Jnk

3430 | www.pnas.org/cgi/doi/10.1073/pnas.1519473113 Matak et al. Downloaded by guest on October 2, 2021 apoptotic pathways, particularly at 10 wk (Fig. 6 D and G). In sum,

our results suggest that Tfr1-CKO DA neurons attempted to INAUGURAL ARTICLE overcome axonal loss and cellular stress at 3 wk, but by 10 wk they committed to cell death. Discussion DA neurons offered a tractable and well-characterized system to investigate neuronal iron transport and how it might be involved in neurodegeneration. We used CKO mice to examine the roles of Fpn and Tfr1 in vivo. Others had proposed that inhibition of Fpn had a pathogenic role in PD (19), but DA neurons lacking SEE COMMENTARY Fpn did not show evidence of iron overload or dysfunction. Accordingly, many patients with FPN mutations have been identified (44) but have not been reported to have an increased frequency of neurodegenerative diseases. Iron exported by FPN must be oxidized, generally by ceruloplasmin, to be loaded onto extracellular TF (8). Interestingly, patients deficient in ceruloplasmin have increased brain iron and neurological symptoms including parkinsonism (45). Importantly, iron accu- mulates in astrocytes in this disease and, at least early, neurons may be iron-starved, similar to ceruloplasmin-deficient mice (46). This observation suggests that sequestered iron is not available to neurons, even though it is increased locally. Iron insufficiency may also cause neurodegeneration in mice lacking iron regulatory protein 2 (IRP2) (47). Considering these observations suggesting that neuro- Fig. 3. Loss of Tfr1 in DA neurons leads to abnormal behavior. White bars, CTR; nal iron deficiency might be detrimental, we developed CKO mice black bars, Tfr1-CKO. (A) Overall distance traveled by CTR and Tfr1-CKO mice lacking Tfr1 exclusively in DA neurons. over60minintheOF[CTRn = 14; Tfr1-CKO n = 13; t(1,25) = 4.959, P < 0.001]. – – Tfr1-CKO mice failed to thrive after 7 wk and died by 12 wk, (B and C) Akinesia and catalepsy in mice at 5 6and9 10 wk of age, two-way MEDICAL SCIENCES likely from inadequate oral intake similar to dopamine-deficient ANOVA. For akinesia, we observed a significant main effect of age [F(1,44) = 10.410, P < 0.02] and genotype [F(1,44) = 32.568, P < 0.001]; the age by ge- mice (48), although we did not exclude involvement of enteric DA notype interaction was also significant [F(1,44) = 10.413, P < 0.002]. Bonfer- neurons. Although 3-wk Tfr1-CKO mice were similar to CTR mice, roni test showed a significant decrease in steps in Tfr1-CKO mice over time they subsequently developed SNpc neurodegeneration. Dorsal (both P < 0.05). When assessed for catalepsy, we also observed a significant striatal terminals disappeared first, followed by SNpc cell bodies and, = < = < age [F(1,42) 36.137, P 0.001] and genotype effect [F(1,42) 27.55, P 0.001] later, VTA cell bodies. DA neurons elsewhere in the brain appeared = < and age by genotype interaction [F(1,42) 31.064, P 0.001]. Bonferroni test to be spared. Tfr1-CKO mice had decreased dopamine synthesis showed a significant increase in catalepsy in the Tfr1-CKO mice that wors- rates and tissue monoamine levels in the dorsal and ventral striata ened in severity over time (P < 0.001). In both experiments CTR n = 5; Tfr1- CKO n = 6at5–6wkandCTRn = 12; Tfr1-CKO n = 7at9–10 wk. (D) L-DOPA and a heightened response to dopamine receptor agonists including in vivo synthesis rates in the frontal cortex (FC) at 7 wk, hippocampus (HIPP), L-DOPA. Our results suggested abnormalities in presynaptic and ventral striatum (VS), and dorsal striatum (DS). Data are presented as ng/mg/hr postsynaptic dopamine receptor responses consistent with selective (CTR n = 12; Tfr1-CKO n = 14); Student’s t test, ***P < 0.001. (E) Tissue levels nigrostriatal neurodegeneration and “murine parkinsonism” (29). of dopamine in the FC, HIPP, VS, DS, and olfactory bulb (OB). Data are The polysome-associated RNA profile of ventral midbrain Tfr1- presented as ng/mg (CTR n = 14; Tfr1-CKO n = 16); Student’s t test, ***P < CKO DA neurons indicated that they were iron deficient. Preser- 0.001. (F) After habituation to the OF for 1 h, mice were given 5 mg/kg car- vation of Tfr1-CKO DA neurons outside the SNpc suggested that bidopa, returned to the OF for 10 min, administered 20 mg/kg L-DOPA, SN DA neurons were more dependent on Tfr1 for iron uptake or = = andplacedintotheOFfor170min(CTRn 12; Tfr1-CKO n 9). Before were particularly sensitive to iron deficiency, perhaps due to their administration of carbidopa, RMANOVA confirmed a significant effect of time [F = 19.699, P < 0.001], and treatment by genotype in- unusually large metabolic needs (49). At 8 wk, mitochondria in SN (11,209) DA neurons were enlarged and disrupted, similar to those ob- teraction [F(11,209) = 4.347, P < 0.001]. Bonferroni tests found, as expected, that the locomotion for Tfr1-CKO mice was lower than that for CTRs at all served in iron-deficient cardiomyocytes lacking Tfr1 (7). Mito- preinjection time-points (P < 0.008). Post–L-DOPA injection, RMANOVA chondrial autophagy appeared to be ineffective, because abnormal

revealed a significant effect of time [F(29,551) = 5.382, P < 0.006], genotype mitochondria were abundant, even though neurons should elimi- = < = < [F(1,19) 9.894, P 0.005], and time by genotype interaction [F(29,551) 4.766, P nate them promptly (50, 51). At the same age, VTA DA mito- 0.001]. Bonferroni post hoc comparison confirmed that Tfr1-CKO mice dra- chondria appeared normal, suggesting that they suffered less matically increased their locomotor responses to L-DOPA beginning at 100 min damage or abnormal mitochondrial were effectively removed. We < – and lasting until 220 min postinjection (P 0.043). (G I) OF locomotor activity observed proteinaceous inclusions in processes of Tfr1-CKO DA following i.p. injection of dopamine receptor agonists. (G) 3 mg/kg apomorphine (CTR n = 14; Tfr1-CKO n = 13). A two-way ANOVA revealed the effects of neurons, similar to those reported in DA neurons lacking autophagy protein Atg7 (52) or mitochondrial transcription factor Tfam (53). genotype [F(1,50) = 8.288, P < 0.006] and treatment [F(1,50) = 11.717, P < 0.01] and the treatment by genotype interaction [F(1,50) = 8.375, P < 0.001] to be signifi- We speculate that the inclusions might have arisen from damaged cant. Bonferroni tests showed that apomorphine exerted a pronounced effect and aggregated mitochondria that were not cleared by mitophagy. on Tfr1-CKO activity (P < 0.001). (H) 2.5 or 5 mg/kg SKF-81297 (CTR n ≥ 5; Tfr1- The most strongly up-regulated genes from Tfr1-CKO DA CKO n ≥ 8 across all treatments). A two-way ANOVA for SKF-81297 determined neurons were associated with axonal injury, consistent with distal = < = the effects of genotype [F(1,51) 8.717, P 0.005] and treatment [F(2,51) 3.561, to proximal degeneration, possibly due to metabolic failure in P < 0.036]; genotype by treatment interaction was not significant. Bonferroni long processes. There was marked up-regulation of Pfkfb3, which corrections showed that locomotor responses to 2.5 and 5 mg/kg SKF-81297 is normally inhibited in neurons, and Hk2, which commits cells to were greatly stimulated in Tfr1-CKO mice (ps < 0.05). (I) 0.1 or 0.5 mg/kg quinpirole (D2 receptor agonist; CTR n ≥ 6; Tfr1-CKO n ≥ 8 across all treat- glycolysis. However, neurons are unable to rely on glycolysis, and ments). Two-way ANOVA discerned only the genotype by treatment interaction oxidative stress and neurodegeneration ensue (40, 54). Poly-

to be significant [F(2,51) = 5.522, P < 0.007]. Tfr1-CKO mice showed a potentiated some-associated mRNAs encoding components of all UPR response to 0.5 mg/kg quinpirole relative to CTR animals (P < 0.001). All data pathways were up-regulated in 3-wk Tfr1-CKO DA neurons, are presented as means ± SEM; ***P < 0.001, CTR versus Tfr1-CKO mice. consistent with a cytoprotective response. However, by 10 wk, it

Matak et al. PNAS | March 29, 2016 | vol. 113 | no. 13 | 3431 Downloaded by guest on October 2, 2021 with altered dendrite structure and delayed GABAergic matu- ration, but the authors did not report an increase in cell death (10). Mice lacking IRP2 developed lower motor neuron degeneration, raising the possibility that a lack of utilizable iron can cause the death of some types of neurons (47). However, other investigators have seen less severe (61) or no neurological phe- notypes (62) in independent IRP2 KO strains. Thus, we could not conclude from the literature that iron deficiency causes neuronal cell death. Although our results indicate that loss of Tfr1 caused iron deficiency in ventral midbrain DA neurons, iron deficiency per se may not lead to the death of SNpc DA neurons. It remains possible that some other function of Tfr1 might also be involved. We (5) and others (63) have recently described roles of Tfr1 that do not involve iron uptake. However, transgenic expression of a missense mutant form of Tfr1 that cannot bind Tf (5, 64) did not rescue Tfr1-CKO mice (Fig. S7), indicating that the ability of Tfr1 to take up Tf is necessary for DA neuron survival. Taking into account reports suggesting that iron deficiency predisposes to human PD (20–22) and that iron overload is pro- tective (23–25), we speculate that iron deficiency in DA neurons might contribute to PD pathogenesis. Functional iron deficiency in SNpc neurons could result from impaired iron uptake, as in our mice, or from trapping of iron in damaged mitochondria or protein aggregates, making it unavailable for use. Cellular iron deficiency can lead to loss of activity of oxidative phosphorylation complexes I to IV in cells that are highly dependent on mitochondrial respiration (7), and impaired oxidative phosphorylation may contribute to PD (65). Oxidative stress, de- fective mitochondrial quality control, impaired proteostasis, and the UPR, all observed in our mice, also appear to be important in PD. Interestingly, TFR1 has been identified in computational interaction networks for PD (66, 67).

Fig. 4. Age-progressive neurodegeneration. (A) Representative images from dorsal striatum of Tfr1-CKO mice (3, 5, 7, and 10 wk of age). (B) Rep- resentative SN images from 3, 5, 7, and 10 wk CTR and Tfr1-CKO mice. (C) Numbers of Th+ neurons (Upper) and Nissl+ cells (Lower) in the SNpc at 3, 5, 7, and 10 wk. SNpc cells were quantified with StereoInvestigator by immunolabeling with anti-Th and counterstaining with Nissl (n ≥ 4 for each genotype and timepoint); for Th+ neurons, two-way ANOVA indicated a

significant genotype [F(1,38) = 35.910, P < 0.01] and age [F(3,38) = 6.812, P < 0.01] effect, whereas Bonferroni showed a highly significant decrease in Tfr1-CKO mice at 5, 7, and 10 wk compared with CTRs (P < 0.02). For Nissl+

neurons, two-way ANOVA indicated a significant genotype [F(1,38) = 13.682, P < 0.01] and age [F(3,38) = 9.067, P < 0.01] effect. Bonferroni showed a trend toward a decrease in Tfr1-CKO mice relative to CTRs at 5 wk and a highly significant decrease at 7 and 10 wk (P < 0.02). (D) Representative sagittal sections from 10 wk CTR and Tfr1-CKO mice immunolabeled with anti-Th antibody. NA, nucleus accumbens; OB, olfactory bulb; OTu, olfactory tubercule; SN, substantia nigra; Str, striatum; VTA, ventral tegmental area. All quantitative data are presented as means ± SEM. (Scale bars: A, 2 mm; B, 200 μm; and D, 100 μm.)

appeared that Tfr1-CKO DA neurons had committed to cell death, likely due to prolonged ER and oxidative stress. We observed increased mRNA encoding BH3-only proteins activated downstream of Jnk and Jun in the initiation of neuronal apoptosis (55). This pathway is induced by axonal damage (56). We also observed increased p53, Puma, and Pmaip1/Noxa mRNAs, consistent with induction of that pathway (57), increased Foxo1, encoding a proapoptotic transcription factor implicated in neuronal loss (58), and Casp3, an effector of DA neuron death (59). Fig. 5. Abnormal mitochondria in Tfr1-CKO DA neurons. Representative EM Redd1/Ddit4, which is increased in postmortem samples from images of CTR (A and D) and Tfr1-CKO DA neurons (B, C, E, and F)inthe PD patients (60), was also increased at 10 wk in Tfr1-CKO SNpc. DA neurons are distinguished by Dat immunogold labeling (black dots). Examples of mitochondria are indicated by red arrows. (G and H) + − + − DA neurons. Thus, cell death appeared to be characteristic Representative images of CTR;YFP / and Tfr1-CKO;YFP / DA neurons from of neurodegeneration. 8 wk mice. Green, GFP; blue, DAPI. (Magnification: 200×.) Insets show en- In a previous study, murine hippocampal neurons expressing largement for detail. (I) Blue arrow, example of a proteinaceous inclusion. dominant negative Tfr1 had markedly decreased iron associated (Scale bars: A–F, 500 nm; G,25μm; H,50μm; and I, 100 nm.)

3432 | www.pnas.org/cgi/doi/10.1073/pnas.1519473113 Matak et al. Downloaded by guest on October 2, 2021 INAUGURAL ARTICLE SEE COMMENTARY MEDICAL SCIENCES

Fig. 6. Molecular signatures of neurodegeneration. (A–E) Polysome-associated mRNA levels from microarray analysis from mice at 3 and 10 wk. At 3 wk CTR n = 4, Tfr1-CKO n = 3; at 10 wk CTR n = 3, Tfr1-CKO n = 4. Microarray data are represented in bar graph format as calculated fold-change for Tfr1-CKO relative to CTR. Changes of P < 0.05 were regarded as significant with a fold-change threshold of 2. Results are organized by biological pathway; (F) relative mRNA levels measured by QPCR of additional genes encoding components of ER/UPR pathways at 3 and 10 wk of age; gene identifications are shown at the bottom; (G) relative mRNA levels measured by QPCR of Trp53 and Puma at 10 wk. For F and G, at 3 wk CTR n = 4, Tfr1-CKO n = 3; at 10 wk CTR n = 3, Tfr1-CKO n = 4. Data are presented as means ± SEM; Student’s t test, *P < 0.05, **P < 0.01, ***P < 0.001.

Exposure to high levels of manganese causes a disorder similar deficiency (75). RAB5B, which interacts with PD protein to PD (68). Manganese can substitute for iron in holo-TF (69), LRRK2, is also involved in endosomal trafficking of TFR1 and neurons can take up Mn-TF through receptor-mediated (76, 77). endocytosis (70). Manganese perturbs cellular iron homeostasis, Tfr1-CKO mice, lacking Tfr1 exclusively in DA neurons, rep- resulting in an iron-deficient cellular phenotype (71), but it is not resent an extreme situation that would not occur naturally. known whether this contributes to the clinical picture of However, insights from our studies may be relevant to PD and manganism. other neurodegenerative diseases characterized by disrupted iron Considering that TFR1 is essential for the survival of and iron homeostasis. Decreased activity of TFR1 may contribute to neu- assimilation into DA neurons, one might expect to observe var- rodegeneration in Huntington’s disease (78). Creuztfeldt-Jakob iants in TFR1 in patients with DA neurodegeneration. However, disease was recently reported to involve neuronal iron deficiency, severe loss-of-function mutations are likely to be lethal before apparently due to sequestration of the iron storage protein ferritin birth (3) and have not been described in humans. In- in prion aggregates (79). Additional work will be needed to better terestingly, SNCA and DNAJC13, mutated in patients with understand the possible role of iron deficiency in neurodegener- PD, modulate TFR1 trafficking (72–74). VPS35,anotherPD ative disorders. However, our results already suggest that thera- disease gene, helps sort TFR1 to recycling endosomes, and peutic chelation strategies should be used with caution in PD until insufficiency of its retromer complex causes cellular iron their consequences are better understood.

Matak et al. PNAS | March 29, 2016 | vol. 113 | no. 13 | 3433 Downloaded by guest on October 2, 2021 Materials and Methods glass slides and fixed with DAPI-containing Fluoro Shield solution (Calbiochem). × Animals. Experiments were performed on F2 mice of similar mixed back- Images were taken on a Leica SP5 confocal microscope at 200 magnification. grounds. Floxed 129Sv Fpn (27) or Tfr1 (5) mice were crossed with C57BL/6 DatIRESCre mice (26) to generate Fpn-CKO and Tfr1-CKO mice. Littermates Electron Microscopy and Immunogold Labeling. Anesthetized mice were were used to control for the mixed background. No differences were seen transcardially perfused with 1 g/kg of diethyl-dithiocarbamate followed by between sexes, and both were used. Fpn-CKO and Tfr1-CKO mice were born 5% (vol/vol) glutaraldehyde [25% (wt/vol) stock] and 0.4% sodium meta- in Mendelian ratios. Ribotag transgenic mice (B6N.129-Rpl22tm1.1Psam/J; stock bisulfite in 0.1 M sodium phosphate buffer. Brains were postfixed in 2% (vol/vol) no: 011029) (28) were obtained from Jackson Laboratory. mtYFP mice were paraformaldehyde [32% (wt/vol) stock] and sectioned on a vibratome at kindly provided by Nils-Göran Larsson (Max Planck Institute, Cologne, Ger- 50 nm. Sections were treated with 1% sodium borohydride and blocked with many) (32). The Duke Institutional Animal Care and Use Committee ap- 3% (vol/vol) normal goat serum. Sections were treated overnight at 4 °Cwith proved all animal studies. rat anti-DAT antibody (Millipore) and processed by a pre-embedding immunogold-silver method. Sections were prepared for electron microscopy using Stereology. Immunolabeling with anti-Th or anti-Ddc antibody and cresyl violet standard methods (82), examined on an FEI Morgagni transmission electron (Nissl) counterstaining was performed on every fourth 30-μm section through the microscope, and photographed using an XP-60 digital camera (Advanced SNpc [stereotaxic coordinates Bregma, −3.88 to −2.54 mm (80)] as previously Microscopy Techniques). Detailed methods are described in SI Materials described (81). At least seven tissue sections were counted per mouse. Four and Methods. tissue sections were used to count DA neurons in the VTA (Bregma, −3.52 to −3.08 mm). Four or more mice were used per genotype at each time point. qPCR. cDNA was prepared using High Capacity cDNA RT (Applied Biosciences) Unbiased stereological analysis was performed using the optical fractionator method from 20 to 50 ng RNA either from input or enriched samples. qPCR was with Stereo Investigator software 11 (MBF Bioscience) on a Zeiss AxioImager2 performed using SYBR green (IQ SYBR green Supermix; Biorad) on a Biorad microscope. Counting was performed at 63× by an investigator blinded to C1000 Touch Thermal Cycler using primers shown in Table S3. mRNA ex- genotypes. Additional parameters are detailed in SI Materials and Methods. pression was calculated using the ΔΔCt method and normalized to β-actin.

Immunoprecipitation of Polysomes from DA Neurons. Immunoprecipitation of Microarray Analysis. RNA was extracted using an RNA isolation kit (Qiagen) polysomes from DA neurons in ventral midbrain homogenates was performed according to the manufacturer’s instructions. Samples were treated with DNase I as previously described (28) and as detailed in SI Materials and Methods.We (Qiagen) to remove genomic DNA. Microarray analysis was performed using used 20–50 ng enriched RNA for microarray analysis and 20 ng for cDNA Mouse Expression Array 430A 2.0 (Affymetrix). RNA quality control, hybridiza- synthesis for qPCR analysis. tion, and data analysis were performed by the Duke University Microarray Shared Resource. Data were analyzed using Partek Genomics Suite 6.5 (Partek). Immunohistochemistry. Anesthetized mice were transcardially perfused with Robust multichip analysis (RMA) normalization was done on the entire dataset. 0.1 M PBS/4% (vol/vol) paraformaldehyde [32% (wt/vol) stock]. Tissues were Multiway ANOVA and fold-change were assessed to identify genes differentially + + − postfixed for 12–16 h and cryoprotected in 30% (wt/vol) sucrose-PBS. Frozen expressed between CTR-Ribotag (Tfr1fl/ CRE; Rpl22 / ) and Tfr1-CKO Ribotag + − tissue sections were cut serially to 30-μm thickness. Free-floating sections were (Tfr1fl/fl CRE; Rpl22 / ) mice. Differentially expressed genes were selected with a washed with TBS (0.1 M Tris HCl, pH 7.4, and 0.1 M NaCl). To block endoge- P value cutoff of 0.05 based on the ANOVA test and a fold-change cutoff of 2. nous peroxidase activity, tissues were washed with TBS/10% (vol/vol) hydrogen The full microarray dataset has been deposited online (NCBI accession no. peroxide/10% (vol/vol) methanol and blocked in PBS/5% (vol/vol) normal goat GSE66730). serum/0.25% Triton X-100. Sections were then incubated for 48 h with anti-Th antibody, (1:1,000 striatum, 1:2,000 SN; Calbiotech 657012). Following in- Statistical Analysis. Data are presented as means ± SEM and were analyzed cubation, the sections were washed twice with TBS, incubated for 2 h with with GraphPad Prism 6 for Student’s t test and by repeated-measures biotinylated anti-rabbit secondary antibody and ABC reagent (Vecta Shield Kit; ANOVA (RMANOVA) using the IBM SPSS Statistics programs. Further details Vector Labs), and then washed three times in TBS before development with are provided in Table S1. A log-rank test was used for survival curve com- DAB and mounting. Nissl staining was performed using standard procedures. parisons. P < 0.05 was considered significant. The density of Th immunostaining in the striatum was quantified using ImageJ (rsbweb.nih.gov/ij/). Every fourth section across the striatum from Bregma 1.18 ACKNOWLEDGMENTS. We thank Ramona Rodriguez for help with behav- to 0.70 mm was processed. At least four mice were used per genotype at each ioral data analysis and statistics; the Duke Microarray Core facility (sup- time point. Images were taken on a Zeiss Axio Imager microscope at the Duke ported by National Institutes of Health Grant P30 CA014236) for microarray Microscopy Core Facility. data ascertainment, management, and analysis; Neil Medvitz for electron For immunofluorescence imaging in Fig. 5 G and H, free-floating sections microscopy; Kingshuk Roy Choudhury for statistical analysis; Nils-Göran were blocked for 1 h at room temperature in 10% (vol/vol) donkey serum and Larsson for mtYFP mice; Tomasa Barrientos, Kristina Roberts, Karin Finberg, and Blake Stagg for assistance early in this project; Miquel Vila and 0.1% Triton X-100 and then incubated with anti-GFP antibody (1:400; Calbio- Celine Perier for reviewing electron micrographs; Jim McNamara and Serge chem) for 2 h at room temperature. Sections were then washed with a PBS-0.1% Przedborski for reviewing the manuscript; and the N.C.A. laboratory for Triton X-100 solution for 5 min and subjected to three 5-min washes with PBS, helpful discussions. Some equipment and software was purchased with a followed by 1-h incubation (1:1,000) at room temperature with Alexa 488 sec- grant from the North Carolina Biotechnology Center. This work was sup- ondary antibody (Jackson Immuno Research Labs). Tissues were mounted on ported by National Institutes of Health Grant R01 DK089705 (to N.C.A.).

1. Crichton RR, Dexter DT, Ward RJ (2011) Brain iron metabolism and its perturbation in 10. Fretham SJ, et al. (2012) Temporal manipulation of transferrin-receptor-1-dependent neurological diseases. J Neural Transm (Vienna) 118(3):301–314. iron uptake identifies a sensitive period in mouse hippocampal neurodevelopment. 2. Snyder AM, Connor JR (2009) Iron, the substantia nigra and related neurological Hippocampus 22(8):1691–1702. disorders. Biochim Biophys Acta 1790(7):606–614. 11. Hill JM, Ruff MR, Weber RJ, Pert CB (1985) Transferrin receptors in rat brain: Neu- 3. Levy JE, Jin O, Fujiwara Y, Kuo F, Andrews NC (1999) Transferrin receptor is nec- ropeptide-like pattern and relationship to iron distribution. Proc Natl Acad Sci USA essary for development of erythrocytes and the nervous system. Nat Genet 21(4): 82(13):4553–4557. 396–399. 12. Hentze MW, Muckenthaler MU, Galy B, Camaschella C (2010) Two to tango: Regu- 4. Ned RM, Swat W, Andrews NC (2003) Transferrin receptor 1 is differentially required lation of mammalian iron metabolism. Cell 142(1):24–38. in lymphocyte development. Blood 102(10):3711–3718. 13. Schneider SA, Bhatia KP (2012) Syndromes of neurodegeneration with brain iron 5. Chen AC, Donovan A, Ned-Sykes R, Andrews NC (2015) Noncanonical role of trans- accumulation. Semin Pediatr Neurol 19(2):57–66. ferrin receptor 1 is essential for intestinal homeostasis. Proc Natl Acad Sci USA 112(37): 14. Dexter DT, et al. (1989) Increased nigral iron content and alterations in other metal 11714–11719. ions occurring in brain in Parkinson’s disease. J Neurochem 52(6):1830–1836. 6. Barrientos T, et al. (2015) Metabolic catastrophe in mice lacking transferrin receptor 15. Gorell JM, et al. (1995) Increased iron-related MRI contrast in the substantia nigra in in muscle. EBioMedicine 2(11):1705–1717. Parkinson’s disease. Neurology 45(6):1138–1143. 7. Xu W, et al. (2015) Lethal cardiomyopathy in mice lacking transferrin receptor in the 16. Riederer P, et al. (1989) Transition metals, ferritin, glutathione, and ascorbic acid in heart. Cell Reports 13(3):533–545. parkinsonian brains. J Neurochem 52(2):515–520. 8. Hentze MW, Muckenthaler MU, Andrews NC (2004) Balancing acts: Molecular control 17. Morris CM, Edwardson JA (1994) Iron histochemistry of the substantia nigra in Par- of mammalian iron metabolism. Cell 117(3):285–297. kinson’s disease. Neurodegeneration 3(4):277–282. 9. Rouault TA (2013) Iron metabolism in the CNS: Implications for neurodegenerative 18. Oakley AE, et al. (2007) Individual dopaminergic neurons show raised iron levels in diseases. Nat Rev Neurosci 14(8):551–564. Parkinson disease. Neurology 68(21):1820–1825.

3434 | www.pnas.org/cgi/doi/10.1073/pnas.1519473113 Matak et al. Downloaded by guest on October 2, 2021 19. Lei P, et al. (2012) Tau deficiency induces parkinsonism with dementia by impairing 53. Ekstrand MI, et al. (2007) Progressive parkinsonism in mice with respiratory-chain- APP-mediated iron export. Nat Med 18(2):291–295. deficient dopamine neurons. Proc Natl Acad Sci USA 104(4):1325–1330. 20. Savica R, et al. (2009) Anemia or low hemoglobin levels preceding Parkinson disease: 54. Bolaños JP, Almeida A, Moncada S (2010) Glycolysis: A bioenergetic or a survival INAUGURAL ARTICLE A case-control study. Neurology 73(17):1381–1387. pathway? Trends Biochem Sci 35(3):145–149. 21. Logroscino G, Chen H, Wing A, Ascherio A (2006) Blood donations, iron stores, and 55. Young JE, et al. (2009) Polyglutamine-expanded androgen receptor truncation risk of Parkinson’s disease. Mov Disord 21(6):835–838. fragments activate a Bax-dependent apoptotic cascade mediated by DP5/Hrk. 22. Mariani S, et al. (2016) Association between sex, systemic iron variation and proba- J Neurosci 29(7):1987–1997. bility of Parkinson’s disease. Int J Neurosci 126(4):354–360. 56. Fernandes KA, Harder JM, Kim J, Libby RT (2013) JUN regulates early transcriptional 23. Buchanan DD, Silburn PA, Chalk JB, Le Couteur DG, Mellick GD (2002) The Cys282Tyr responses to axonal injury in retinal ganglion cells. Exp Eye Res 112:106–117. polymorphism in the HFE gene in Australian Parkinson’s disease patients. Neurosci 57. Perier C, et al. (2007) Two molecular pathways initiate mitochondria-dependent do- Lett 327(2):91–94. paminergic neurodegeneration in experimental Parkinson’s disease. Proc Natl Acad 24. Pichler I, et al.; PD GWAS Consortium; International Parkinson’s Disease Genomics Sci USA 104(19):8161–8166. Consortium; Wellcome Trust Case Control Consortium 2; Genetics of Iron Status 58. Chuang CL, Lu YN, Wang HC, Chang HY (2014) Genetic dissection reveals that Akt is SEE COMMENTARY Consortium (2013) Serum iron levels and the risk of Parkinson disease: A Mendelian the critical kinase downstream of LRRK2 to phosphorylate and inhibit FOXO1, and randomization study. PLoS Med 10(6):e1001462. promotes neuron survival. Hum Mol Genet 23(21):5649–5658. 25. Xia J, Xu H, Jiang H, Xie J (2015) The association between the C282Y and H63D 59. Yamada M, Kida K, Amutuhaire W, Ichinose F, Kaneki M (2010) Gene disruption of ’ polymorphisms of HFE gene and the risk of Parkinson s disease: A meta-analysis. caspase-3 prevents MPTP-induced Parkinson’s disease in mice. Biochem Biophys Res – Neurosci Lett 595:99 103. Commun 402(2):312–318. 26. Bäckman CM, et al. (2006) Characterization of a mouse strain expressing Cre re- 60. Malagelada C, Jin ZH, Greene LA (2008) RTP801 is induced in Parkinson’s disease and ′ combinase from the 3 untranslated region of the dopamine transporter locus. mediates neuron death by inhibiting Akt phosphorylation/activation. J Neurosci – Genesis 44(8):383 390. 28(53):14363–14371. 27. Donovan A, et al. (2005) The iron exporter ferroportin/Slc40a1 is essential for iron 61. Zumbrennen-Bullough KB, et al. (2014) Abnormal brain iron metabolism in Irp2 de- – homeostasis. Cell Metab 1(3):191 200. ficient mice is associated with mild neurological and behavioral impairments. PLoS 28. Sanz E, et al. (2009) Cell-type-specific isolation of ribosome-associated mRNA from One 9(6):e98072. – complex tissues. Proc Natl Acad Sci USA 106(33):13939 13944. 62. Galy B, et al. (2005) Altered body iron distribution and microcytosis in mice deficient 29. Sotnikova TD, et al. (2005) Dopamine-independent locomotor actions of amphet- in iron regulatory protein 2 (IRP2). Blood 106(7):2580–2589. amines in a novel acute mouse model of Parkinson disease. PLoS Biol 3(8):e271. 63. Senyilmaz D, et al. (2015) Regulation of mitochondrial morphology and function by 30. Rensvold JW, et al. (2013) Complementary RNA and protein profiling identifies iron as stearoylation of TFR1. Nature 525(7567):124–128. – a key regulator of mitochondrial biogenesis. Cell Reports 3(1):237 245. 64. Schmidt PJ, Toran PT, Giannetti AM, Bjorkman PJ, Andrews NC (2008) The transferrin 31. Xu W, Barrientos T, Andrews NC (2013) Iron and copper in mitochondrial diseases. Cell receptor modulates Hfe-dependent regulation of hepcidin expression. Cell Metab Metab 17(3):319–328. 7(3):205–214. 32. Sterky FH, Lee S, Wibom R, Olson L, Larsson NG (2011) Impaired mitochondrial 65. Shoffner JM, Watts RL, Juncos JL, Torroni A, Wallace DC (1991) Mitochondrial oxi- transport and Parkin-independent degeneration of respiratory chain-deficient do- dative phosphorylation defects in Parkinson’s disease. Ann Neurol 30(3):332–339. pamine neurons in vivo. Proc Natl Acad Sci USA 108(31):12937–12942. 66. Liu Y, Koyutürk M, Maxwell S, Zhao Z, Chance MR (2012) Integrative analysis of 33. Mootha VK, et al. (2003) PGC-1alpha-responsive genes involved in oxidative phos-

common neurodegenerative diseases using gene association, interaction networks MEDICAL SCIENCES phorylation are coordinately downregulated in human diabetes. Nat Genet 34(3): and mRNA expression data. AMIA Jt Summits Transl Sci Proc 2012:62–71. 267–273. 67. Rakshit H, Rathi N, Roy D (2014) Construction and analysis of the protein-protein 34. Subramanian A, et al. (2005) Gene set enrichment analysis: A knowledge-based ap- interaction networks based on gene expression profiles of Parkinson’s disease. PLoS proach for interpreting genome-wide expression profiles. Proc Natl Acad Sci USA One 9(8):e103047. 102(43):15545–15550. 68. Racette BA (2014) Manganism in the 21st century: The Hanninen lecture. 35. Laplante M, Sabatini DM (2012) mTOR signaling in growth control and disease. Cell Neurotoxicology 45:201–207. 149(2):274–293. 69. Aisen P, Aasa R, Redfield AG (1969) The chromium, manganese, and cobalt complexes 36. Chung CY, et al. (2005) Cell type-specific gene expression of midbrain dopaminergic of transferrin. J Biol Chem 244(17):4628–4633. neurons reveals molecules involved in their vulnerability and protection. Hum Mol 70. Suárez N, Eriksson H (1993) Receptor-mediated endocytosis of a manganese complex Genet 14(13):1709–1725. of transferrin into neuroblastoma (SHSY5Y) cells in culture. J Neurochem 61(1): 37. Yasuda M, Tanaka Y, Ryu M, Tsuda S, Nakazawa T (2014) RNA sequence reveals 127–131. mouse retinal transcriptome changes early after axonal injury. PLoS One 9(3):e93258. 71. Crooks DR, Ghosh MC, Braun-Sommargren M, Rouault TA, Smith DR (2007) Manga- 38. Kanaan NM, et al. (2015) The longitudinal transcriptomic response of the substantia nese targets m-aconitase and activates iron regulatory protein 2 in AF5 GABAergic nigra to intrastriatal 6-hydroxydopamine reveals significant upregulation of re- cells. J Neurosci Res 85(8):1797–1809. generation-associated genes. PLoS One 10(5):e0127768. 72. Ben Gedalya T, et al. (2009) Alpha-synuclein and polyunsaturated fatty acids promote 39. Herrero-Mendez A, et al. (2009) The bioenergetic and antioxidant status of neurons is – controlled by continuous degradation of a key glycolytic enzyme by APC/C-Cdh1. Nat clathrin-mediated endocytosis and synaptic vesicle recycling. Traffic 10(2):218 234. α Cell Biol 11(6):747–752. 73. Chai YJ, et al. (2013) The secreted oligomeric form of -synuclein affects multiple – 40. Rodriguez-Rodriguez P, Fernandez E, Almeida A, Bolaños JP (2012) Excitotoxic stim- steps of membrane trafficking. FEBS Lett 587(5):452 459. ulus stabilizes PFKFB3 causing pentose-phosphate pathway to glycolysis switch and 74. Fujibayashi A, et al. (2008) Human RME-8 is involved in membrane trafficking through – neurodegeneration. Cell Death Differ 19(10):1582–1589. early endosomes. Cell Struct Funct 33(1):35 50. 41. Hetz C, Mollereau B (2014) Disturbance of endoplasmic reticulum proteostasis in 75. Chen C, et al. (2013) Snx3 regulates recycling of the transferrin receptor and iron – neurodegenerative diseases. Nat Rev Neurosci 15(4):233–249. assimilation. Cell Metab 17(3):343 352. 42. Hauser DN, Hastings TG (2013) Mitochondrial dysfunction and oxidative stress in 76. Bucci C, et al. (1995) Co-operative regulation of endocytosis by three Rab5 isoforms. – Parkinson’s disease and monogenic parkinsonism. Neurobiol Dis 51:35–42. FEBS Lett 366(1):65 71. 43. Silva RM, et al. (2005) CHOP/GADD153 is a mediator of apoptotic death in sub- 77. Trischler M, Stoorvogel W, Ullrich O (1999) Biochemical analysis of distinct Rab5- and stantia nigra dopamine neurons in an in vivo neurotoxin model of parkinsonism. Rab11-positive endosomes along the transferrin pathway. J Cell Sci 112(Pt 24): – J Neurochem 95(4):974–986. 4773 4783. 44. Pietrangelo A, Caleffi A, Corradini E (2011) Non-HFE hepatic iron overload. Semin 78. Lumsden AL, Henshall TL, Dayan S, Lardelli MT, Richards RI (2007) Huntingtin- Liver Dis 31(3):302–318. deficient zebrafish exhibit defects in iron utilization and development. Hum Mol 45. Miyajima H (2015) Aceruloplasminemia. Neuropathology 35(1):83–90. Genet 16(16):1905–1920. 46. Jeong SY, David S (2006) Age-related changes in iron homeostasis and cell death in 79. Singh A, Qing L, Kong Q, Singh N (2012) Change in the characteristics of ferritin in- the cerebellum of ceruloplasmin-deficient mice. J Neurosci 26(38):9810–9819. duces iron imbalance in prion disease affected brains. Neurobiol Dis 45(3):930–938. 47. Jeong SY, et al. (2011) Iron insufficiency compromises motor neurons and their mi- 80. Franklin BJ, Paxinos G (2007) The Mouse Brain in Stereotaxic Coordinates (Elsevier, tochondrial function in Irp2-null mice. PLoS One 6(10):e25404. San Diego), 3rd Ed. 48. Zhou QY, Palmiter RD (1995) Dopamine-deficient mice are severely hypoactive, 81. Benner EJ, et al. (2004) Therapeutic immunization protects dopaminergic neurons in a adipsic, and aphagic. Cell 83(7):1197–1209. mouse model of Parkinson’s disease. Proc Natl Acad Sci USA 101(25):9435–9440. 49. Pacelli C, et al. (2015) Elevated mitochondrial bioenergetics and axonal arborization 82. Sesack SR, Hawrylak VA, Matus C, Guido MA, Levey AI (1998) Dopamine axon vari- size are key contributors to the vulnerability of dopamine neurons. Curr Biol 25(18): cosities in the prelimbic division of the rat prefrontal cortex exhibit sparse immuno- 2349–2360. reactivity for the dopamine transporter. J Neurosci 18(7):2697–2708. 50. Youle RJ, van der Bliek AM (2012) Mitochondrial fission, fusion, and stress. Science 83. Gaier ED, et al. (2013) Peptidylglycine α-amidating monooxygenase heterozygosity 337(6098):1062–1065. alters brain copper handling with region specificity. J Neurochem 127(5):605–619. 51. Batlevi Y, La Spada AR (2011) Mitochondrial autophagy in neural function, neuro- 84. Jones SR, et al. (1998) Profound neuronal plasticity in response to inactivation of the degenerative disease, neuron cell death, and aging. Neurobiol Dis 43(1):46–51. dopamine transporter. Proc Natl Acad Sci USA 95(7):4029–4034. 52. Friedman LG, et al. (2012) Disrupted autophagy leads to dopaminergic axon and 85. Veznedaroglu E, Milner TA (1992) Elimination of artifactual labeling of hippocampal dendrite degeneration and promotes presynaptic accumulation of α-synuclein and mossy fibers seen following pre-embedding immunogold-silver technique by pre- LRRK2 in the brain. J Neurosci 32(22):7585–7593. treatment with zinc chelator. Microsc Res Tech 23(1):100–101.

Matak et al. PNAS | March 29, 2016 | vol. 113 | no. 13 | 3435 Downloaded by guest on October 2, 2021